Multivariate CART for survival with dental applications

描述（由申请人提供）：牙周病或龋齿导致的牙齿缺失困扰着大多数成年人的一生。专业牙科治疗和个人口腔健康自我护理的目标是防止牙齿脱落并保持自然牙列处于舒适和功能状态。为了制定适当的牙科治疗计划，牙科保健专业人员必须权衡保留每颗牙齿的长期前景与每种治疗方案所涉及的成本和不适，患者和第三方付款人必须对建议的牙科治疗的成本做出价值判断。因此，在制定任何牙科治疗计划时，都必须利用牙科预后。目标：本研究的主要目标是将新开发的分类和回归树（CART）的多变量扩展应用于从Creighton University School of Dentistry的AxiUm电子牙科记录系统获得的大型去识别数据库，以建立基于实际牙齿脱落的全球牙齿预后指标。我们还计划改进我们为生存多变量CART开发的计算机算法，以提高计算效率，并进一步将贝叶斯生存树和随机森林扩展到相关结果，以便利用最准确的程序建立牙齿预后分配标准。这些假设将在一个为期4年的纵向数据库中进行测试，该数据库包括19，092名在Creighton University牙科诊所接受治疗的男性，女性和儿童的随访数据。在19,092名有随访数据的患者中，4,339名成年人是牙周诊所的常规患者，有随访牙周记录，而1,681名是18岁以下的儿童。用于检验假设的可用数据：年龄，性别，某些记录中的种族，牙齿脱落的时间，完整的修复和牙周记录以及随访，重要的病史，吸烟，口腔卫生，特殊需求，依从性，牙科就诊频率和其他参数的数据，这些数据是在Creighton University School of Dentistry进行牙科治疗和随访时收集的。重要性：通过建立明确的，客观的标准，分配准确的，基于证据的全球牙科预后指标，这些指标是基于实际的牙齿脱落，这项研究有可能对牙科社区产生巨大的影响。采用和使用准确的、基于证据的全球牙科预后指标可以帮助牙科医生更好地制定治疗计划，帮助第三方支付计划更好地做出有关福利的决定，并为患者提供一个就其牙科治疗需求做出明智决定的途径。

DESCRIPTION (provided by applicant): Tooth loss from periodontal disease or dental caries afflicts most adults over the course of their lives. Goals of professional dental treatment and of personal oral health self-care are to prevent tooth loss and maintain the natural dentition in a state of comfort and function. In order to develop appropriate dental treatment planning, dental health care professionals must weigh the long-term outlook for retention of each tooth against the cost and discomfort involved with each treatment option, and patients and third-party payers must make value judgments about the cost of suggested dental treatments against the probability of long-term success for those treatments. Hence, it is essential to utilize dental prognosis in the development of any dental treatment plan. Goals: The primary goal of this study is to apply newly developed multivariate extensions of Classification And Regression Trees (CART) for survival to a large, de-identified database obtained from the AxiUm system of electronic dental records at Creighton University School of Dentistry to establish global dental prognostic indicators based on actual tooth loss. We also plan to refine computer algorithms we developed for multivariate CART for survival to increase computing efficiency and further extend Bayesian survival trees and random forests to correlated outcomes in order to utilize the most accurate procedure for establishing criteria for assignment of dental prognosis Subjects: The hypotheses will be tested on a 4-year longitudinal database consisting of 19,092 men, women, and children with follow-up data who have been treated at the Creighton University dental clinic. Of the 19,092 patients with follow-up data, 4,339 adults are regular patients in the periodontal clinic with follow-up periodontal records while 1,681 are children under 18 years of age. Available data for testing hypotheses: Data for age, gender, race in some records, time until tooth loss, complete restorative and periodontal records with follow-up, significant medical history, smoking, oral hygiene, special needs, compliance, frequency of dental visits and other parameters that are collected as part of dental treatment and follow-up at the Creighton University School of Dentistry. Significance: By establishing clear, objective criteria for assignment of accurate, evidence-based global dental prognostic indicators that are based on actual tooth loss, this study has the potential to make a tremendous impact on the dental community. Adoption and use of accurate, evidence-based global dental prognostic indicators could aid dental practitioners in better treatment planning, help third-party payment plans make better decisions about benefits, and provide patients with an avenue for making informed decisions about their dental treatment needs.